JOURNAL OF OPTOELECTRONICS AND ADVANCED MATERIALS Vol. 10, No. 5, May 2008, p.

1192 - 1196

Comparison between model and experiment in studying
the electric arc


S. NITU
*
, C. NITU, C. MIHALACHE, P. ANGHELITA
a
, D. PAVELESCU
University POLITEHNICA from Bucharest, 313 Spl. Independentei, 060032 Bucharest, Romania
a
Research and Development Institute for the Electrotechnic Industry, 313 Spl. Unirii, 030138 Bucharest, Romania



The paper presents a comparison between the results obtained using different modeling equations for the electric arc
(Habedank, KEMA, Schavemaker and Schwarz) and the experimental measurements, in order to determine the best of
them for predicting the behavior of a low voltage vacuum circuit breaker.

(Received March 13, 2008; accepted May 5, 2008)

Keywords: Electric arc model, Vacuum interrupter, Post-arc current, Transient recovery voltage



1. Introduction

The switching tests, to which are subjected the
interrupters, consume a large amount of power and
materials. Thats why efforts are made to obtain an ante-
calculation of the arc voltage, the post-arc current and the
transient recovery voltage, the electric parameters
definitory for the switching process. There are some
equations (Mayr, Cassie and derived from) modeling the
switching process, by taking into account the apparatus
characteristics, such as dissipated power and the time
constants.
Within the framework of this paper, the most suitable
model of the switching phenomena in an a.c. low voltage
vacuum circuit breaker is looked for, aiming to better
predict the apparatus behaviour. From numerous and very
attentive experimental investigations of the switching
processes taking place at high currents interruption,
statistically processed, the parameters of the electric arc
model used in the modeling equations are determined.
Finaly, the arc voltage, the post-arc current and the
transient recovery voltage, which represent the circuit
breaker stresses in the switching process, are calculated.
The obtained results are compared with the experimental
ones in order to establish the best modeling equation for
the studied low voltage vacuum circuit breaker, able to
predict the behaviour of other similar circuit breakers or to
optimise an existing apparatus.


2. Modeling equations for the electric arc
dynamic regime

At the beginning of the last century, A.M. Cassie and
O. Mayr have done the first tests (1930) concerning this
subject. The equation belonging to Mayr is the most used
up to now for investigation of the dynamic regime of the
electric arc. Mayr model [1] is applied for the
approximation of the electric arc behavior in the range of
current passing through zero and is described by the
equation:
|
|
.
|

\
|

= 1
1
d
d 1
0
P
i u
t
G
G
m
t
(1)

where: u and i instant values of the arc voltage and
current respectively; G dynamic conductance of the
electric arc column supposed to have an exponential
variation versus the stored energy,
m
time constant of
the electric arc at current zero moment, P
0
- dissipated
power at current passing through zero, considered to be
constant.
Cassie model [1] represents a good approximation for
the high current electric arc; he considered a linear
variation of the column arc conductance with the stored
energy and modeled the arc dynamic behavior with the
following equation:

|
|
.
|

\
|
= 1
1
d
d 1
2
2
a c
U
u
t
G
G t
(2)

where are used the same symbols like in (1) with
c
time
constant of the electric arc at the maximum current value
and U
a
- the electric arc voltage, considered to be
constant.
The other known equations or systems of equations
for the electric arc modeling (Habedank, KEMA,
Schavemaker and Schwarz) are derived from Mayr and
Cassie models and contain four or more parameters. The
main problem of the electric arc simulation by means of
any model is to determine the values of the parameters,
which are not really constants, due to the high speed
evolution of the physical phenomena in the extinguishing
chambers of the interrupters. Although the parameters are
considered constants and they are calculated using the
oscillograms resulted from experiments. The determined
parameters are valid just for a certain type of interrupter,
Comparison between model and experiment in studying the electric arc

1193
characterized by specific features of burning and cooling
of the electric arc.
The four parameters of Mayr and Cassie models are
also the parameters of the Habedank model [2] which
considers the electric arc conductance (G=

i/u) being a
Cassie conductance (G
c
=

i/u
c
) and a Mayr conductivity
(G
m
=

i/u
m
) in series. It is a system with three unknowns:
G
c
, G
m
and G and contains the Cassie and Mayr model
equations, modified, because:
c c
G G u u = and
m m
G G u u = . The system is the following:

(
(

|
|
.
|

\
|
= 1
d
d
2
2
2
c a c
c c
G
G
U
u G
t
G
t
(3)
|
|
.
|

\
|
= 1
d
d
0
2
m c
m m
G
G
P
G u G
t
G
t
(4)
m c
G G G
1 1 1
+ =
(5)

Schwarz model [3], derived from Mayr model,
considers an exponential variation with the electric arc
conductance of (G

) and P
0
(P
0
G

), where and P
0

are the initial values, at the current zero moment; the
constants and are obtained by iteration, until calculated
and experimental results are well fitting. Kema model [3]
is formed from three Mayr models, that is from 3
conductances (G
i
) in series, each one calculated from a
modified Mayr equation; each conductance has an
exponential variation
i
n
i
G , but with different values of the
exponents ( ) 2 , 1 e
i
n . Schavemaker model [4] reunites the
Cassie and Mayr models in the same equation.


3. Calculated and experimental results

Fig.1 and 2 represent typical oscillograms for the
interrupting process in a low voltage vacuum circuit
breaker, obtained in an experimental setup, able to provide
asymmetric short-circuit currents up to 54 kA
rms
(110
kAmax) at 805 or 1360 V
rms
and to perform accurate and
repeatabil determinations.



Fig.1. Typical oscillogram with i
pa
- the post-arc current,
i
sc
the short circuit current and u
a
the arc voltage.


Fig. 2. The transient recovery voltage (trv) for the
interruption from Fig.1.; U=1140V is the maximum
value of the voltage supply (805V
rms
).

The short circuit interrupted current (i
sc
) has a
maximum value of 24 kA and has an asymmetrical
behavior, due to the testing circuit parameters. The arc
voltage (u
a
) has an average value of U
a
=

20V and the post-
arc current (i
pa
) a maximum value of 7.8A. The transient
recovery voltage has a frequency of f
0
=

23.5kHz and an
over oscillation factor =1.4.
For determining the constants P
0
and
m
, necessary for
the Mayr model, it was used the Asturian method [5],
which starts from experimental results of measured current
and voltage, in order to calculate numerically the arc
conductance and its derivative, at equally distanced time
intervals, t. The estimation of those values of P
0
and
m
,
which lead to the smallest difference between calculated
and experimental conductance derivatives is made by the
least squares method [6]. This way it is obtained a pair of
values
m
and P
0
which can be used for modeling the
electric arc in the investigated device. Because two
oscillograms are not identical, even the experimental
conditions are conserved for the same switching device,
several pairs
m
and P
0
were determined. These ones can
be obtained for the same device, but it is preferably for
several devices from the same series. Average values of
the pairs
m
and P
0
should be used in the model, in order to
provide more general results. In the Cassie arc model
equation, U
a
is the average value of the arc voltage, so
only the time constant
c
must be numerically calculated,
in a similar way as described for P
0
and
m
.
From the experimental data presented in Fig.1 and 2
the parameters obtained [6] for the Mayr and Cassie
models, so that the modeled phenomena fit as well as
possible the experiment are: P
0
=15.75kW;
m
=3.177s;
U
a
=20V;
c
=4s. The initial conditions, necessary to solve
the differential equation are: G(0)=9700S and u(0)=0.
In order to obtain the 4 interesting variables for the
interrupting process (the interrupted short circuit current,
the post-arc current, the arc voltage and the transient
recovery voltage), are used the MATLAB Simulink and
the Arc Model Blockset [3]. The arc models, with their
specific parameters, are implemented in a circuit,
calculated to provide a short circuit current and a transient
S. Nitu, C. Nitu, C. Mihalache, P. Anghelita, D. Pavelescu

1194
recovery voltage similar with those obtained
experimentally. Using the Mayr model, the voltage arc is
very close to zero; using the Cassie model, no circuit
breaking occurs (Fig.3). Thats why the other models,
which combine Mayr and Cassie models, are further
investigated.





Fig.3. Comparison between Mayr (M) and Cassie (C) arc
models.


The Habedank model has the advantage of using the 4
parameters already calculated for Mayr and Cassie
models. These 4 parameters are mathematically calculated
and the problem has a unique solution. For the other
models: KEMA, Schwarz and Schavemaker, the
parameters are determined by successive trials, and the
solution is not unique. So, the Habedank model is the one
with which the comparison of the other models is made.
The results are synthetic presented in the Fig. 4, 5 and
6. These figures are made by reuniting the modellization
results for the same time scale: a half period or the current
zero moment. There are also different scales for voltages
and currents, on the same axe.



Fig. 4.a. Comparison between Habedank (H) and KEMA
(K) arc models: the arc voltage and interrupted current


The Habedank model is better because the arc voltage
has an almost constant value of 21V and the short circuit
current is longer, closer to the experimental measurements
(11ms). In the current zero moment, the Habedank model
is also better, because, for almost similar recovery voltages
(f
0
=

24kHz, =1.57), the KEMA model does not calculate
any post-arc current, while the Habedank model calculates
a post arc current of 8A, similar with the experimental one.




Frig.4.b. Comparison between Habedank (H) and KEMA
(K) arc models in current zero moment: the arc voltage,
the transient recovery voltage and post-arc current


The KEMA arc model parameters used to solve the
equations system from [3] are: T1=3.177s; A1=0.8 10
-4

1/W; A2=10.5 10
-5
1/W; k1=10; k2=20; k3=80 with the
same initial conditions: G(0)=9700S and u(0)=0.
The Schwarz model leads to calculated results also
worst then Habedank model: an arc voltage variation far
from the experimental one, with very big value before
current zero moment (1150V), a transient recovery voltage
with higher oscillation factor (

=1.76) and a very short
post-arc current, three times shorter than the one
calculated with Habedank model. Also the Habedank
model calculates a post arc current shorter than the
experimental one.




Fig.5.a. Comparison between Habedank (H) and Schwarz (B)
arc models: the arc voltage and interrupted current
Comparison between model and experiment in studying the electric arc

1195


Fig.5.b. Comparison between Habedank (H) and
Schwarz (B) arc models in current zero moment: the arc
voltage, the transient recovery voltage and post-arc
current.


The Schwarz model parameters used to solve the
equation [3] are: =40s; =0.5373; P
0
=100MW; =0.3,
with the same initial conditions G(0)=9700S and u(0)=0.
The parameters are determined by successive trials. The
important parameters are and , because they determine
the current and voltage variation. For fixed value of and
, the other 2 parameters, P
0
and , can vary in a rather
large domain.
The Schavemaker arc model gives good results: the
arc voltage remains almost constant, with little peak before
current zero moment and the short circuit current is a little
longer then the one calculated with Habedank model. The
transient recovery voltage is almost the same, at both
compared models, but the post arc current is very short
and has a maximum value of only 2.75A calculated with
the Schavemaker model.
The parameters used to solve the Schavemaker arc
model equation [3] are: =1.227s; Ua=20V; P0=2750W;
P1=0.55 with the same initial conditions G(0)=9700S and
u(0)=0.



Fig.6.a. Comparison between Habedank (H) and Schavemaker
(S) arc models: the arc voltage and interrupted current



Fig.6.b. Comparison between Habedank (H) and Schavemaker
(S) arc models in current zero moment: the arc voltage, the
transient recovery voltage and post-arc current

4. Conclusions

The errors between the Habedank model and
experiment are around 30% concerning the post arc
current duration and transient recovery voltage amplitude,
but only 5% concerning the amplitude of the post arc
current and the frequency of the transient recovery
voltage. The errors introduced by the method of
m
,
c
and
P
0
calculation are in the same range (5%). The calculated
curves have a faster variation, the post arc current and the
oscillations of the transient recovery voltage have a shorter
duration.
Thats why a model which takes into account also the
thermal effect of the interrupted current combining the
Mayr and Cassie models, as Habedank and Schavemaker
models, will lead to better results. The Habedank model
presents the advantage of determining the 4 parameters by
calculation starting from experimental oscillograms. The
method of calculation leads to a unique solution. For
Schavemaker model, only Ua is calculated as the average
value of the measured arc voltage. The other three
parameters must be obtained by successive trials and the
solution is not unique.
So the Habedank model is considered to give the best
results concerning the modellization of the studied vacuum
circuit breaker behavior. It is important to obtain, further, a
better interpretation of the parameters model and their
relation with the vacuum circuit breaker design.


References

[1] G. Hortopan, Aparate electrice de comutatie, vol
1and 2, Editura Tehnica Bucuresti, 2000
[2] U. Habedank, EtzArchiv, 11, 339 (1988).
[3] *** Arc Model Blockset
S. Nitu, C. Nitu, C. Mihalache, P. Anghelita, D. Pavelescu

1196
http://www.ewi.tudelft.nl/live/pagina.jsp?
id=8abc1adb-2254-4a34-a8ad-
84bb89049459&lang=en
[4] P. H. Schavemaker, L. van der Sluis, Proc. Of the
Second IASTED International Conf. Power and
Energy Systems, Greece, p 644, 2002
[5] W. Gimenez, O. Hevia, Apendice III: Articulos
Publicados en Congresos Internacionales, 2000,
Universidad Tecnologica Nacional, Fac. Reg. Santa
Fe, Argentina, pp.41-45.
[6] S. Nitu, C. Nitu, P. Anghelita, The International
Conference on Computer as a Tool, EUROCON 2005.
Belgrade, Serbia and Montenegro 2, 1442,
ISBN: 1-4244-0049-X


__________________________
*
Corresponding author: scnitu@rdslink.ro or
snitu@apel.apar.pub.ro